Proinsulin endopeptidase substrate specificities defined by site-directed mutagenesis of proinsulin.

Two endopeptidases are involved in the conversion of proinsulin; a type I activity directed at the B chain, Arg31,Arg32, C-peptide junction, and type II which cleaves the C-peptide, Lys64,Arg65, A chain junction. To define further the substrate specificities of these enzymes, a series of mutant preproinsulin cDNAs were generated by site-directed and deletion mutagenesis. These were inserted into pT7 plasmids and capped cRNA transcripts synthesized, that were then microinjected into Xenopus oocytes. Oocytes were biosynthetically radiolabeled with [3H]leucine and the secreted peptides (greater than 95% present as unprocessed proinsulins) then incubated with types I and II endopeptidase activities prepared from isolated insulinoma secretory granules. The reaction products were analyzed by high performance liquid chromatography. Des-38-62-proinsulin, in which all but six amino acids of C-peptide were deleted was not processed by either enzyme. The mutant Lys64,Arg65 to Thr64,Arg65 was not cleaved by the type II enzyme but was still a substrate for the type I enzyme. The mutant Arg31,Arg32 to Arg31,Gly32 correspondingly was not cleaved by the type I enzyme; however, in this case it was not attacked by the type II enzyme. These results indicate that not only is the presence of a dibasic sequence essential, but also that the secondary structure of the protein is important in determining whether the prohormone is susceptible to proteolytic processing.

Two endopeptidases are involved in the conversion of proinsulin; a type I activity directed at the B chain, Ar8',ArgS2, C-peptide junction, and type I1 which cleaves the C-peptide, LysB4,Args6, A chain junction.
To define further the substrate specificities of these enzymes, a series of mutant preproinsulin cDNAs were generated by site-directed and deletion mutagenesis. These were inserted into pT7 plasmids and capped cRNA transcripts synthesized, that were then microinjected into Xenopus oocytes. Oocytes were biosynthetically radiolabeled with ['Hlleucine and the secreted peptides (>95% present as unprocessed proinsulins) then incubated with types I and I1 endopeptidase activities prepared from isolated insulinoma secretory granules. The reaction products were analyzed by high performance liquid chromatography.
Des-38-62proinsulin, in which all but six amino acids of C-peptide were deleted was not processed by either enzyme. The mutant Lyss4,Args6 to E a 4 , A r g e 6 was not cleaved by the type I1 enzyme but was still a substrate for the type I enzyme. The mutant Arg'',Arg'' to ArgS1,mS2 correspondingly was not cleaved by the type I enzyme; however, in this case it was not attacked by the type I1 enzyme. These results indicate that not only is the presence of a dibasic sequence essential, but also that the secondary structure of the protein is important in determining whether the prohormone is susceptible to proteolytic processing.
The intracellular conversion of proinsulin is thought to involve the concerted activity of three proteinases; two endoproteases (type I and type 11) (Davidson et al., 1988) and an exopeptidase (carboxypeptidase H) (Docherty and Hutton, 1983;Davidson and Hutton, 1987). The type I endoprotease cleaves exclusively on the C-terminal side of Ar$l,Ar$' (B chain/(=-peptide junction), whereas the type I1 endoprotease cleaves preferentially on the C-terminal side of L y~~* , A r g~~ (C-peptide/A chain junction) (Davidson et al., 1988). Both endopeptidases display an acidic pH optimum and are activated by calcium. Studies with group-specific proteinase inhibitors have failed to define the mechanism of catalysis and the possibility remains that they constitute a new class * This work was supported in part by the Medical Research Council, British Diabetic Association, Wellcome Trust, and Nordisk Insulinlaboratorium. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
To whom correspondence should be addressed.
11 Juvenile Diabetes Foundation Research Fellow. of proteinase. Active site-directed reagents have shown that the primary amino acid sequence of the dibasic site is an important determinant of endoprotease substrate specificity. Thus, the type I activity was more susceptible to inhibition by an Arg-Arg containing sulfonium salt than by a Lys-Arg analogue, and the converse was true for the type I1 activity (Rhodes et al., 1989).
Cleavage of peptides at sites marked by pairs of basic amino acids is a common feature of prohormone processing (Docherty and Steiner, 1982). In many cases multiple copies of a bioactive peptide (e.g. met-enkephalin) or different bioactive peptides (e.g. proopiomelanocortin) may reside in the same precursor sequence, permitting tissue-specific processing to generate different peptides from the same parent molecule (Loh et al., 1984). Not all paired basic sequences are cleaved within a propolypeptide, suggesting that factors in addition to the primary sequence are involved in determining the specificity of processing endoproteases.
By site-directed mutagenesis of the encoding DNA, it is now possible to change specific amino acid residues in a protein (for reviews, see Knowles (1987) and Shaw (1987)) and thereby examine the importance of a particular amino acid, sequence, or structural motif on catalytic function of enzymes or the susceptibility of protein substrate to enzymic attack. To examine further the substrate specificity of the proinsulin processing type I and type I1 endoproteases, we have used oligodeoxynucleotide site-directed mutagenesis to change the paired basic sites within proinsulin. Mutant radiolabeled proinsulins for use as substrates were synthesized by microinjecting appropriate cRNAs into Xenopus oocytes.

MATERIALS AND METHODS
Animals-Xenopus oocytes were purchased from Xenopus Ltd., Redhill, Surrey, United Kingdom.
Oligodeoxynucleotide Site-directed Mutagenesis-oligodeoxynucleotide site-directed mutagenesis was performed as described by Carter ' The abbreviations used are: dATPaS, deoxyadenosine 5 ' -~r - [~~S ] thiotriphosphate; SDS, sodium dodecyl sulfate; BisTris, 2[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)-propane-l,3-diol; HPLC, high performance liquid chromatography. et al. (1985). The human preproinsulin cDNA was subcloned into the EcoRI site of M13K19 replicative form DNA and transfected into competent Escherichia coli strain TG1. Single stranded template was prepared, and mutant second strand synthesized using the Klenow fragment of DNA polymerase in the presence of T4 DNA ligase. Two primers were used simultaneously: one a mutagenic oligodeoxynucleotide which generated the desired nucleotide change, and a second selection oligodeoxynucleotide (Sel 2) which converted an EcoK site to an EcoB site. The double stranded DNA was purified from a low melting point agarose gel and used to transfect a nonsuppressor, repair-deficient E. coli strain, HB2154. The resultant plaques were transferred using a sterile toothpick to an asymmetric grid on an Lplate and grown up as colonies of infected bacteria for 15 h. A nitrocellulose blot was prepared and probed with "P-labeled mutagenic oligodeoxynucleotides. Filters were subjected to autoradiography after washing a t two levels of stringency ( 3 X 10-min washes in 6 X SSC a t room temp and 3 X 10 min washes in 6 X SSC at 60 "C where 1 X SSC is 0.15 M NaC1, 0.015 M sodium citrate). Positive colonies, i.e. those which produced a signal after the high stringency wash, were obtained with a frequency of approximately 50% of the colonies which were positive with the low stringency wash. Single stranded template was prepared from positives and sequenced by the dideoxy chain termination method (Sanger et al., 1977) using Sequenase (Tabor and Richardson, 1987) and [cu-3sS]dATPcuS.
T o mutate an additional site in the same cDNA, the mutagenesis reaction was performed on the single stranded mutant template with a second mutagenic primer and a selection primer (Sel 3) which converted the EcoB site to an EcoK site. The double stranded DNA was purified as above and used to transfect E. coli HB2155.
I n Vitro Transcription-The normal and mutant cDNAs were subcloned into the EcoRI site of plasmid pT71 and capped cRNA synthesized from Hind111 linearized plasmid in a reaction containing 40 mM Tris-HCI (pH 8.0), 15 mM MgC12, 5 mM dithiothreitol, 1 mM each of ATP, CTP, and UTP, 0.1 mM GTP, 0.5 mM miG(5')ppp(5')G, 0.25 mg/ml bovine serum albumin (RNase-and DNase-free), 10 units RNaseguard, 70 units of T, RNA polymerase, and 1 pg of plasmid DNA. The mixture was incubated a t 37 "C for 60 min. After treatment with DNase (RNase-free) a t 37 "C for 10 min, the cRNA was extracted once with pheno1:chloroform:isoamyl alcohol (25:24:1) and once with ch1oroform:isoamyl alcohol (24:l). The cRNA was then precipitated using 0.1 volumes 7 M ammonium acetate and 2.5 volumes ethanol. The pellet was collected following centrifugation for 10 min a t 12,000 X g and the cRNA resuspended in water a t a concentration of 1-2 r g l d .
Assay of Proinsulin Conversion-Insulin secretory granules were prepared from rat transplantable insulinomas as described previously (Davidson et al., 1988) and lysed in a buffer comprised of 20 mM BisTris, 0.5% (v/v) Triton X-100,0.5 mM EDTA, 10 p~ phenylmethylsulfonyl fluoride, 10 p M pepstatin A, 10 pM trans-epoxysucciny1-Lleucylamido-(4-guanidino)butane (E-64), and 50 p~ tosylphenylalanylchloromethane (TPCK) (pH 5.5). Insoluble material was removed from the lysate by centrifugation and subjected to anion exchange chromatography on a DE-52 resin, as described previously (Davidson et al., 1988). to separate the type I and type I1 endopeptidase activities. In the standard assay 20 pl of the incubation medium derived from five radiolabeled oocytes (10,000-25,000 dpm of trichloroacetic acidprecipitable material) was added to a 1.5-ml capacity microcentrifuge tube and made to a final volume of 100 pl in a buffer containing (final concentrations) 50 mM sodium acetate (pH 5.5), 1 mM MgCI,, 5 mM CaCI,, 10 p M phenylmethylsulfonyl fluoride, 10 pM pepstatin A, 10 pM E-64, and 50 p~ TPCK, carboxypeptidase B (50 pg/ml), and 10-30 pg of protein containing the enzymic activities. The incubation was terminated after 3 h a t 30 "C by addition of 300 pl of ice-cold 5% (w/v) trichloroacetic acid. A mixture of human proinsulin, des 31,32 proinsulin, des 64,65 proinsulin, and insulin (2-5 pg each in 10 pI 10 mM HCI) was added at this stage as carrier and to serve as calibration standards for HPLC. The mixture was centrifuged for 5 min a t 12,000 X g, the pellet washed further with 1 ml of 5% (w/v) trichloroacetic acid and then extracted twice with 500 pl of diethyl ether. After airdrying a t room temperature, the precipitated material was reconstituted in 120 pl of 10 mM HCI, a 10-pl sample removed for determination of total radioactivity and 100 pl then loaded onto a 30 X 0.8cm column packed with Lichrosorb C-18 silica resin (Merck, Darmstadt, West Germany). The column was developed a t 1 ml/min as described previously (Davidson et al., 1988), and 0.6-ml fractions collected. T o each of these was added 0.5 ml of water and 4 ml of Optiphase Hisafe I1 scintillation mixture (LKB, Milton Keynes, United Kingdom) and their radioactivity determined by liquid scintillation counting.

RESULTS
Site-directed Mutagenesis-Oligodeoxynucleotide site-directed mutagenesis was used to change the coding sequence for the paired basic residue processing sites within proinsulin. The following oligodeoxynucleotides were designed (position of mismatch is indicated by asterisk after corresponding nucleotide): D34 (5' TCTGCCTCCCC*GCGGGTC 3') for converting Arg31,Arg32 t o Arg31,&32, and D35 (5' ATGCCACGCG*TCTGCAGG 3') for converting L y~~~, A r g~~ to Thr64,Arg65. The changes were confirmed by DNA sequencing as shown in Fig. 1. The double mutant A r $ * , W combined with m64,Arg6s was also generated using oligodeoxynucleotides D34 and D35.
In addition to the above, a mutant cDNA containing a deletion of the major part of the C-peptide coding region (des-38-62-proinsulin) was generated by using strategically situated restriction enzyme sites (Shennan and Docherty, 1988).

WT W T D35 A C G T A C G T A C G T A C G T A B
FIG. 1. Sequence of wild-type and mutant preproinsulin cDNAS. A, DNA sequence encoding the Arg",Ar$* processing site of the wild-type preproinsulin ( W T ) and Ar$',%'* site of the mutant preproinsulin ( 0 3 4 ) generated using oligodeoxynucleotide D34. The normal cDNA has the sequence CCCGG and the mutant the sequence CCCCG as indicated by the arrow. B, DNA sequence encoding the LysM,Arg6s processing site of the wild-type preproinsulin ( W T ) and -Thr64,Arg65 site of the mutant preproinsulin ( 0 3 5 ) generated using oligodeoxynucleotide D35. The normal cDNA has the sequence CGCTT and the mutant the sequence CGCGT as indicated by the arrow.
Oocytes-A major radiolabeled protein of M; 9,000 was present in the media of oocytes microinjected with wild-type preproinsulin cRNA or with the D34-, D35-, and D34/35-generated mutants, but not in the media from water-injected oocytes (Fig. 2). This protein migrated ahead of the preproinsulin marker (MI ll,OOO), indicating that the prepeptide had been removed and that normal or mutant proinsulins were secreted into the media. Scanning densitometry of the fluorographs confirmed that the radiolabeled proinsulin represented >95% of the labeled proteins secreted from the oocytes. The expression and secretion of des-38-62-proinsulin was less efficient than the wild-type or other mutant proinsulins. This is due to the effect of the deletion on the stability of the mRNA in microinjected Xenopus oocytes (Shennan and Docherty, 1988) and on the rate of secretion of the mutant proinsulin (Shakur et al., 1989).
In Vitro Processing of Mutant Proinsulins-Xenopus oocytes microinjected with cRNA corresponding to the wildtype proinsulin cDNA secreted a radiolabeled peptide with a retention on HPLC identical to a human recombinant proinsulin standard (Fig. 3, panel 1). No other major components were evident in the secreted product. The majority of the trichloroacetic acid-precipitable radioactivity loaded onto the column was recovered in the eluted fractions (Table I), further substantiating the impression gained from the electrophoretic analysis ( Fig. 2) that the proinsulin molecule accounted for a high proportion of the secreted radioactivity. It was thus evident that the pre-sequence of the preproinsulin expressed in the oocytes had been removed by co-translational proteolysis and that little or no endogenous proteolytic processing at paired basic sites had occurred. It also appeared unlikely that the oocytes produced any other post-translational modifications to the proinsulin molecule, since these would probably have affected the chromatographic behavior of the molecule.
Partial proteolysis of the secreted material with insulin secretory granule lysates in the presence or absence of car- of wild-type and mutant proinsulins in Xenopus oocytes. Xenopus oocytes (10) were microinjected with 50 nl of wild-type or mutant preproinsulin cRNA and incubated at 20 "C for 6 h in modified Barth's solution. The oocytes were then incubated in 30 ~1 of modified Barth's solution containing 30 &i of [3H]leucine for 18 h. The media was removed, the volume made up to 100 pl with modified Barth's saline, and a 5~1 aliquot analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. The samples are: "C-labeled molecular weight markers (track 1) and 13Hlleucinelabeled preproinsulin which was synthesized in a cell-free translation system (truck 2) (see Shennan and Docherty, 1988); the media from oocytes microinjected with, water (track 3), wild-type preproinsulin cRNA (track 4), and mutant cRNAs generated with oligodeoxynucleotide D34 (track 5), D35 (track 6), and a combination of D34 and D35 (track 7). boxypeptidase H activity produced peptides corresponding to cleavage of proinsulin at the paired basic residues. Incubation of the expressed wild-type proinsulin with the insulin granule type I endopeptidase activity in the presence of carboxypeptidase H yielded des-31,32-proinsulin or, in its absence, split 32,33-proinsulin (Fig. 3, panel 1). Similarly, incubation with the type II activity in the presence of carboxypeptidase H yielded des-64,65proinsulin and lesser quantities of insulin and des-31,32-proinsulin. In the absence of carboxypeptidase H, the reaction products were identified as the corresponding peptides extended with basic amino acids. These results were entirely consistent with the previously reported specificities of these enzymes determined using iz51-proinsulin as substrate (Davidson et al., 1988).
With the exception of des-38-62-proinsulin, the mutant preproinsulin genes were expressed with efficiencies approximating that of the wild-type gene as indicated by similar quantities of trichloroacetic acid-precipitable material in the oocyte media, In each case a single major peak of radioactivity was detected. There was little evidence for the endogenous processing of any of the mutant proinsulin molecules by the oocytes before their secretion into the media or of the secretion of other endogenous peptides in the microinjected cells.
Assays of the conversion of the mutant proinsulins were conducted using conditions where the conversion of the wildtype proinsulin obeyed first-order kinetics. Incubations were terminated at a point when around 50-70% of the wild-type proinsulin would have been converted by the type I and type II endopeptidases (Fig. 3, panel 1). This protocol allowed assessment of the effects of the mutations on the rates of proteolysis and moreover increased the probability of detection of any intermediate forms in the reaction.
The mutant des-38-62-proinsulin was not cleaved by either type I or type II endopeptidase activity either alone or in combination (Fig. 3, panel 2).
The mutant Thr64,Arg65 proinsulin was converted by the type I activity to a single major radioactive product which eluted slightly before the wild-type split 32,33-proinsulin (Fig.  3, panel 3; Table I). When carboxypeptidase H was also included in the incubation, there was again a single radioactive product which eluted in a different position slightly before the wild-type des-31,32-proinsulin. Since this mutant proinsulin also elutes slightly before the wild-type proinsulin, it is concluded that the molecule has been cleaved by the type I activity in the expected manner after the unmodified Arg3',Arti2 sequence. The m",Arg6" proinsulin mutant, unlike the wild-type proinsulin, was not cleaved significantly by the type II activity which normally cleaves preferentially on the C-terminal side of the LysB4,Arge sequence. Processing in the presence of the type I and II endopeptidases was equivalent to the processing observed in the presence of type I activity alone.
Mutation of the A@,Arg?* sequence to A$',w* prevented cleavage by the type I endopeptidase activity which is directed at this site (Fig. 3, panel 4; Table I). However, surprisingly, this mutant was a poor substrate for the type II endopeptidase in spite of there being no alteration in the normal Lysa,ArgG sequence in this mutant. Little processing was likewise observed in the combined presence of type I and II activities or with unfractionated secretory granule lysates, The double mutant, Ar$i,w combined with ma,ArgG was also poorly processed by either enzyme (data not shown), an observation which was expected on the basis of the above findings with the singly mutated molecules.

TABLE I
The proteolytic processing of wild-type and mutant proinsulins by insulin secretory granule lysates The radioactive proteins in oocyte incubation media were incubated in the presence of secretory granule lysates, precipitated with 5% (w/v) trichloroacetic acid, and analyzed by HPLC (see "Materials and Methods"). The retention of the respective proinsulin, the percentage of the original protein radioactivity recovered from the column, and the extent of conversion of the precursor is indicated. The number of analyses (n) performed are shown. ACN, acetonitrile.

DISCUSSION
The use of mutant proinsulins generated by site-directed mutagenesis shows that the proinsulin processing type I and type I1 endoproteases are highly specific for sequences containing pairs of basic amino acids. The type I enzyme activity, which is specific for Ar$l,Ar$' did not cleave at this site when the sequence was changed to Ar$',GlP, whereas the type I1 enzyme activity, which is directed principally at the L y~& , A r g~~ site did not cleave at this site when the sequence was changed to Thr64,Arg65.
In addition, the results also demonstrated that the second-ary structure of proinsulin was important in defining the activity of the two endoproteases. When the L y~~, A r g~~ cleavage site was changed to ThP4,Arg6', both the type I and the Arg-Arg-specific component of the type I1 enzyme cleaved at the Ar$l,Ar$' site. However, when the Argl,Ar$' cleavage site was changed to Ar$l,Gl~~~, the type I1 enzyme did not cleave efficiently at the LysM,Arg5 site. One would expect this result if the processing of proinsulin were sequential, i.e. that the Lys-Arg site is susceptible to cleavage only after initial cleavage at the Arg-Arg site. However, the fact that the type I1 activity can produce split 65,66-proinsulin from wildtype proinsulin argues against this. There is similarly no indication that the split 32,33-proinsulin is preferred to proinsulin by the type I1 activity as a substrate, since processing in the combined presence of type I and I1 activities would have exceeded the sum of the activities where incubated alone. It is more likely that the activity of the enzymes may be dependent on the secondary structure around the cleavage site and that this structural requirement was disrupted by changing the Ar$l,Ar$' site to Ar$',Gly3'. The conclusion that the structure around the dibasic cleavage sites is important is supported by the observed inability of the type I and type I1 enzymes to cleave at the dibasic sites in des-38-62proinsulin.
It is difficult to predict what effect specific changes introduced into the proinsulin molecule would have on the overall secondary structure, since no three-dimensional x-ray crystallographic data on proinsulin (or any other prohormone) is available. Secondary structure predictions using the algorithms of Chou and Fasman (1978) indicated that the point mutations at the dibasic cleavage sites do not have any effect on the a helical or p sheet structure of proinsulin. The cleavage sites, it is predicted, are located in or close to p turns and are flanked by an a helix or p sheet. The same structure is present at the dibasic sites of des-38-62-proinsulin except that a predominant /3 turn in the central portion of the Cpeptide is no longer present, and the mini C-peptide is in a p sheet. Rholam and Cohen (1986) have made a survey using the Chou and Fasman rules of dibasic sites within a number of prohormones, and have concluded that dibasic processing sites may be located within p turns, whereas those dibasic sites which are not cleaved are located in a helices or p sheets.
However, in our mutants where the dibasic site is not cleaved, i.e. the L y~~~, A r g~~ site in Gly3' proinsulin, and the two dibasic sites in des-38-62-proinsulin, there is no indication that these sites are present in a helix or p sheet structures.
Our results showing that the type I or type I1 endoproteases were unable to cleave des-38-62-proinsulin contrasts with the findings of Thim et al. (1986), who found that des-38-62proinsulin and proinsulins with C-peptides only two amino acids long were cleaved at one or other of the dibasic sites when expressed in yeast. This could be due to differences in specificity between the proinsulin processing endoproteases and the yeast endoprotease (presumably KEX2 (Julius et al., 1983)) responsible for processing expressed prohormones.